Evaluation of Sentinel Lymph Nodes in Complex Atypical Endometrial Hyperplasia

1. Introduction

Complex atypical hyperplasia (CAH) of the endometrium is a precursor lesion for endometrial cancer. Up to 42% of cases of CAH diagnosed on endometrial biopsy are found to harbor endometrial cancer on hysterectomy specimens [1]. Guidelines for the management of complex endometrial hyperplasia lack clear direction [2] and practice patterns vary from no staging, intraoperative endometrial assessment, and conventional lymph node sampling [3,4,5,6]. This lack of consistency ultimately results in some endometrial cancer cases being un-staged, undertreated, or overtreated [7].

The American College of Obstetrics and Gynecology Committee currently recommends a total abdominal hysterectomy, with or without bilateral salpingo-oophorectomy, including an option for intraoperative evaluation for possible staging of an endometrial cancer [2].

Most of those cancers are low risk [8] defined as grade 1-2 endometrioid adenocarcinomas that are less than or equal to 2 cm in size that invade less than or equal to 50% of the myometrium. Those cancers have been found to have a low risk of Lymph node metastasis and therefore do not need staging by lymph node dissection [9,10].Nevertheless, evaluating these factors by frozen section analysis has been found to lack validity in most institutions, as evidenced by low accuracy rates reported in various studies [11,12,13]. This poses a challenge, as the procedure requires time and resources that are not universally available [14].

Sentinel lymph node sampling is now an established staging procedure for endometrial cancer¹⁵ that obviates the need for comprehensive lymph node dissection with its attendant potential morbidity and cost [16,17,18,19]. The validity of this technique has now been well documented in large studies [20,21] and an NCCN guideline algorithm exists that informs its use in endometrial cancer [22].

Laparoscopic staging of endometrial cancer by identifying sentinel lymph nodes under near-infrared light after injecting Indocyanine Green (ICG) dye in the cervix has a high map-rate and negative predictive value and its safety has been well documented [15,16,20,21].

The use of sentinel lymph node sampling in CAH stems from studies that have demonstrated that up to 10% of cases were classified as high-risk for metastatic disease [1,2,3,4,5,6,7,8], with a risk of lymph node metastasis of 3-7% [8]. Sentinel lymph node sampling in CAH had a high map-rate and negative predictive value, and without adding morbidity or surgical time [7,8,23].

The aim of this study is to evaluate the efficacy and safety of sentinel lymph node sampling in cases of complex atypical endometrial hyperplasia surgically managed by the da Vinci (Intuitive Surgical, Sunnyvale CA) robotic platform using cervical injection of ICG dye under near-infrared light.

2. Results

One-hundred and thirteen (N=113) patients were included in the study. Sixty-nine patients underwent Sentinel lymph node sampling (SLN), and 44 did not (NO SLN). The average age in the SLN cohort was 63 years, and in the NO SLN cohort was 61.2 years. The median BMI was significantly lower in the SLN cohort (34.3 Kg/m²) compared to the NO SLN cohort (40.0 Kg/m²) (p= 0.004). Twenty-six percent of the cohort had diabetes mellitus, 65% had hypertension, and 34% had hyperlipidemia with no statistical difference between the two sub-cohorts. Most patients had an ASA grade of 2-3 with no statistical difference between the two cohorts (Table 1).

Table 2 summarizes the histopathological variables of our cohort. Sixty-three (55.8%) patients had their diagnosis through hysteroscopy dilation and curettage, and fifty (44%) patients were diagnosed by endometrial biopsy (p=0.44). Thirty-nine (34.5%) patients were diagnosed with complex atypical hyperplasia of the endometrium either bordering on endometrial cancer or cannot rule out endometrial cancer (CAH/EAC) (p= 1.00). Fifty-two (46%) patients were diagnosed with endometrial cancer on final pathology, 32 (46.4%) in the SLN cohort and 20 (45.5%) in the NO SLN cohort (p= 0.215).

Table 3 describes the characteristics of patients with endometrial cancer. Fifty-two (46%) patients were diagnosed with endometrial cancer on post operative histopathological analysis, 32 were in the SLN cohort and 20 were in the NO SLN cohort (Table 3). All patients had endometrioid histopathology, there were no cases of uterine carcinosarcoma, uterine serous cancer or clear cell carcinoma of the endometrium. Fifty (96.2%) patients had FIGO grade 1 or 2 disease, and 2 (3.8%) had grade 3 disease. There was no statistical difference between the two cohorts.

Thirty-six (69%) patients were diagnosed with stage 1A cancer, and 15 (28.8%) had stage 1B. There was no statistical difference between the two cohorts. Only one patient -in the SLN cohort- had stage II cancer. One patient with stage IB disease had positive isolated tumor cells (ITC) LN metastasis. According to consensus guidelines [15] this did not up-stage the patient to stage III C1, and the patient was treated with adjuvant brachytherapy to the vaginal cuff following NCCN guidelines²² for treatment of stage IB disease (Table 3).

Thirty-six (69.2%) patients had a depth of invasion less than or equal to 50%, and 16 (30.8%) patients had the depth of invasion more than 50%. There was no statistical difference between the two cohorts. The size of the tumor was less than or equal to 2 centimeters in 19 (36.5%) patients, and more than 2 centimeters in 27 (51.9%) patients. The size of the tumor was not determined in 6 (11.5%) patients. There was no statistical difference between the two cohorts. Lymphovascular space invasion was reported in 3 (5.8%) patients in the SLN cohort. Peritoneal cytology was not reported to be positive for malignant cells in any patient (Table 3).

There was no statistical difference between the two cohorts regarding the preoperative diagnosis or the biopsy method with 29 (55.8%) patients diagnosed with complex atypical hyperplasia bordering on or cannot rule out cancer (CAH/EAC), and 26 (50%) patients diagnosed by curettage (D&C) (Table 3).

Twenty seven out of 52 cancer patients did not meet low-risk Mayo Clinic criteria (grade 1-2; less than or equal to 2 cm in greatest diameter; and depth of myometrial invasion less than or equal to 50%). These patients comprised 23.8 % of the entire cohort and would have required lymph node assessment [9,10].

On multivariate analysis, only preoperative histology (complex atypical hyperplasia bordering on or cannot rule out cancer (CAH/EAC) vs. complex atypical hyperplasia (CAH)) was found to be a significant predictor of cancer (P<0.001). Age, BMI, Diabetes mellitus, hypertension, hyperlipidemia, and biopsy method did not predict cancer in our model (Appendix A).

Thirty-nine (34.5%) patients were diagnosed with CAH bordering on EAC (CAH/EAC) preoperatively. Twenty-nine out of thirty-nine (74.3%) were diagnosed with cancer postoperatively versus 23/74 (31%) in those preoperatively diagnosed with CAH (P<0.001) (Table 3). There was no statistical difference between those two subgroups regarding age, BMI, diabetes mellitus, hypertension, hyperlipidemia, biopsy method, postoperative grade or stage, size of the tumor or depth of invasion (Appendix B). Multiple logistic regression analysis using the same variables demonstrated no significant difference between the two subgroups (Appendix C).

Bilateral mapping to the sentinel lymph nodes (Table 4) was achieved in 64 (92.8%) of the 69 patients in the SLN cohort while unilateral mapping was achieved in the remaining 5 (7.2%) patients. A lymph node was recovered from both hemi-pelvises in 63/64 bilaterally mapped patients and from all patients that mapped unilaterally (Table 4). Pelvic lymph node dissection was performed in 4 patients in the SLN cohort and in 4 patients in the NO SLN cohort, while para-aortic lymph node dissection was performed in 1 patient in the NO SLN cohort. A lymph node was reported to be positive for isolated tumor cells (ITC) LN in 1 patient in the SLN cohort (Table 4).

Operative findings are summarized in Table 5. The mean operative time was 148.8 minutes in the SLN cohort and 144.3 minutes in the NO SLN cohort (p = 0.918). The mean estimated blood loss was 92 milliliters in the SLN cohort and 109 milliliters in the NO SLN cohort (p = 0.009). There was no conversion to laparotomy. Grade 3 or 4 Clavien-Dindo [26,27] complications were reported in no patients in the SLN cohort and 4 (9%) patients in the NO SLN cohort (P = 0.021), with 3 patients with grade-3 and 1 patient with grade-4 complications (Table 5). A detailed narrative of those complications is provided in Appendix D.

3. Discussion

In this single institution retrospective cohort study, the prevalence of endometrial cancer in patients with a preoperative diagnosis of CAH was 46%. Most patients with cancer had a low-grade disease. All of them had early-stage disease, but 23.8% met Mayo criteria for lymph node dissection. When sentinel lymph node sampling was attempted, bilateral mapping was achieved in 92.8% of patients and unilateral mapping was achieved in 7.2%. Sentinel lymph node sampling was not associated with additional complications.

SLN evaluation has become a widely adopted technique for endometrial cancer staging due to its safety and high sensitivity in detecting lymph node metastasis when combined with ultra-staging using immunohistochemistry [20,21,28,29], and an NCCN algorithm is a sanctioned standard of care in its management [22,24]. SLN Mapping using cervical injection of ICG dye under near infrared light has shown an improved sensitivity and negative predictive value [16; 20-21], and although the technology is available in different platforms, it is safe to assume that most gynecological oncologists in the United States are currently well trained to perform the procedure expeditiously and safely using the da Vinci (Intuitive Surgical, Sunnyvale CA) robotic platform [7,20,30,31].

Given that the risk of endometrial cancer is up to 42% [1] in patients with a preoperative diagnosis of CAH investigators have begun to study sentinel lymph node sampling in the management of such patients [5,7,8]. The necessity for this research has stemmed from the fact that the sensitivity of a preoperative biopsy in the diagnosis of cancer is low with a high rate of inter and intra-observer variability on the largest prospective GOG trial [1,32]. Intraoperative diagnosis of cancer by frozen section analysis has been reported in studies as an effective method [9,33], however the technique has not been shown to be reliable outside of a few centers with dedicated personnel, time and resources [11,12,13,14] additionally, sentinel lymph node sampling after a hysterectomy has severed the uterine lymphatics is technically challenging.

It is true, however, that the prevalence of high-intermediate risk cancer or high-risk cancer in this population is about 10% [1,8] and the incidence of lymph node metastasis is about 3-7% [8]. Although identifying those patients would inform adjuvant treatment and prevent either reoperation or overtreatment with radiation therapy [34,35], a surgical technique that benefits such a relatively small number of patients must have an equally low risk of complications.

Retrospective studies on the use of SLNS in CAH have consistently reported a low complication rate [7,36]. We used a validated surgical complication score [26,27] and found no grade 3 or 4 morbidity associated with SLNS. Indeed, all 4 of our grades 3-4 complications were in the NO SLN cohort likely reflecting the statistically unbalanced nature of our retrospective cohort where patients with a higher BMI and more comorbidities did not undergo SLNS.

As mentioned above our study is limited by its retrospective nature and hence the selection bias inherent in our analysis. The small sample size allowed for a limited multiple logistic regression analysis model which, consistent with findings by Touhami et al. [5] identified a preoperative diagnosis of complex atypical hyperplasia, cannot rule out or bordering on cancer (CAH/EAC) as a significant predictive factor of a postoperative diagnosis of cancer.

However, we did not evaluate the contribution of other predictive factors such as preoperative endometrial thickness, which was found to predict cancer where an endometrial thickness of ≥ 15 mm was associated with a 1.5 times risk of cancer and a 2.5 times risk of meeting Mayo clinic criteria for lymph node dissection [37,38]. Furthermore, Laskov et al. [39] reported on 79 patients with CAH where CAH in a polyp was associated with significantly less incidence of not only cancer but also higher grade and stage disease [39]. Including as many relevant factors as possible in a preoperative algorithm to instruct the most efficient management policy is certainly warranted.

Another possible selection bias that appeared in our study is the significantly higher BMI in patients in the NO SLN cohort. This most likely reflects the surgeon’s reluctance to perform a procedure whose risk may outweigh its benefit. Kogan et al. [40] reported on 223 obese (median BMI 40.6 Kg/m²) patients with endometrial cancer where SLN sampling was not associated with a difference in overall survival or progression free survival while being associated with a significant increase in operative time and blood loss although the clinical significance of the small change reported is unclear. Whether obesity is a risk factor for unsuccessful SLN mapping is controversial in the literature [41,42].

Nevertheless, this study elucidated the low sensitivity of preoperative sampling whether by endometrial biopsy or curettage to diagnose cancer in contradistinction to the high reliability and safety of performing sentinel lymph node sampling in this population where 10%-28% meet criteria for lymph node dissection [8,37].

In conclusion, given the low risk of SLNS, we support a prospective study into its clinical benefit in patients with a preoperative diagnosis of CAH. As conceptualized by others, such a study should include a preoperative radiological [38] histopathological [5,39] and molecular [7] classification algorithm of specimens to further the diagnosis and inform the management of this entity.

4. Materials and Methods

This was a retrospective cohort study conducted between December of 2016 and March of 2023 after obtaining Institutional Review Board l (IRB) approval [number 201504071 J]. The study included all patients ranging in age between 18 and 90 years with complex atypical endometrial hyperplasia diagnosed on preoperative biopsy or on a dilation and curettage sample and managed surgically using the da Vinci (Intuitive Surgical, Sunnyvale CA) robotic platform. The authors excluded all patients with hyperplasia of the endometrium without atypia.

This cohort was treated by two surgeons at our institution. All patients underwent a robot (da Vinci platform, Sunnyvale, CA) assisted total laparoscopic hysterectomy and bilateral salpingo-oophorectomy. Sentinel lymph node sampling was conducted when feasible or when deemed necessary by the surgeon based on their interpretation of the preoperative histopathological findings. Complex atypical hyperplasia of the endometrium bordering on endometrial cancer or cannot rule out endometrial cancer (CAH/EAC) was an indication for sentinel lymph node sampling unless contraindicated by other variables such as morbid obesity, serious comorbidities such as coagulopathy or heart disease, or, in some cases, lack of mapping under near-infrared light after injection of indocyanine green (ICG) dye in the cervix. The da Vinci platform-mounted near-infrared light camera was used for the identification of sentinel lymph nodes and ICG dye (0.5mg/ml) was injected in the cervix in the standard manner reported previously at 3 o’clock and 9 o’clock, 2 millimeters deep and 1 centimeter deep [24]. The National Comprehensive Cancer Network (NCCN) [22] algorithm for sentinel lymph node sampling in endometrial cancer was followed in most cases undergoing SLN dissection. If mapping for SLN was negative in one hemi-pelvis then a systematic lymph node dissection of this hemi-pelvis was performed [22] unless the risk was deemed to outweigh the benefit. Additionally, excision of suspicious lymph nodes and para-aortic lymph node dissection were performed at the surgeon’s discretion. An intraoperative frozen section analysis was inconsistently used, and the findings shall not be reported in this analysis.

The numerical variables were summarized using median and interquartile range (first quartile – third quartile). The categorical variables were summarized using frequencies and percentage. To compare the cohorts, the Student’s t-test or Wilcoxon Rank Sum test was used for numerical variables as appropriate. The chi-square test or Fisher exact test was used for the categorical variables, based on the sample size. Multiple logistic regression analysis was used to assess the chance of having cancer with the variables of interest. The 2-sided p value was reported for each test. A p value less than 0.05 was considered an indication of statistical significance. Statistical analysis was performed using the R language [25].